Using the GUI you can basically switch on and off certain predefined functionality, and set some parameters.
Show Waypoints - If this is checked, the waypoints are displayed. Only currently to an agent assigned waypoints can change their visibility, so you might have to wait a bit before all waypoints appear/disappear. Since several agents share one waypoint, it is not possible to display the details of the waypoint for each agent (they can adapt to the agent's direction). So only the information of the last agent assigned will be displayed, resulting in the waypoints changing their look during the simulation.
Show Quadtree - Displays the quadtree used to store the agents internally. In order to find neighbors of agents quickly, they are grouped together in cells. These cells are generated and arranged dynamically.
Show Direction - Displays the agent's direction. The yellow line represents where they actually do walk at the moment. The length of this line represents the velocity of the agent.
Show Forces - Displays the forces that affect the individual agents. The forces are shown towards the direction the agents is accelerated. Red: direction they would like to walk to (desired direction). Blue: force that pulls them away from walls. Green: force that pulls them away from each other. Magenta: "Look Ahead" force.
Framerate - Specify how many updates per second shoud be made. If the requested value is higher than what your computer can deliver, it will have no effect.
Look Ahead - This mental layer strategy is a bit more sophisticated. Each agent looks ahead a certain distance and counts the other agents to his left, and to his right, respectively. It then walks slightly into the direction where less agents were counted. Only agents in front of the agent, and only those with a walking direction in the opposite direction are considdered. So walking in lines behind each other is not affected.
Dijkstra Routing - Not implemented in this demo (yet?), because the scenario is too small for this feature.
Visual Avoidance -Not implemented in this demo (yet?).
Precision (h) - In each timestep (frame) of the simulation, the agent is allowed to walk forward a tiny step. All the accelerations and velocoties are scaled to match this step length. The precision defines how long such a step is (High precision = small steps). Setting precision too low will allow the angents to walk through walls and each other, since they only detect an obstacle when they are already through it. If the precision is high, the simulation will run very slowly, but the results are not supposed to change dramatically (except for rounding errors in the calculations, which can cause a slightly different overall result). The value h is also known as tau in literature and the documentation of the social force model.
Wall Force - This slider defines how strong the force pushing away from the walls and other obstacles is (fiW). The higher it is, the bigger the distance an agent will keep from the wall will be. If the wall force is too low, agents will be able to walk through the walls. This will be more likely the larger the force between pedestrians becomes (see next slider), especially if the pedestrian density is very high (bottlenecks, or during an unevenly initialization).
Pedestrian Force - Defines the distance each agent tries to keep from the other agents (fij). Set this to a high value if no mental layer strategies are activated (i.e. the simulation is run as a pure social force model simulation). Some mental layer strategies try to steer agents around other agents in advance, to this force is only used as a last change to avoid a collision. (or to actually simulate a collision, if set to a very low value.)
(c) Christian Gloor [ c|h|g|l|o|o|r|@|s|i|l|m|a|r|i|l|.|o|r|g| ]